Ep. 050 - Your Groups Are Too Small | SAMPLE SIZE |
ฝัง
- เผยแพร่เมื่อ 29 ส.ค. 2024
- On this episode, Seth is joined in the studio by Project Engineer Miles Neville, and Senior Ballistician Jayden Quinlan. While this topic can be somewhat controversial, the reality is math doesn’t lie. The guys discuss different ways of load development, velocity ES/SD, and assessing a rifle’s performance using data that is statistically significant. Many precision rifle shooters are looking at data sets that are simply too small to properly represent their rifle’s performance. In this podcast we look to dispel some dogma and see how statistics play a role rifle and load analysis.
Best bit of info in this podcast. 20 shot minimum for establishing true zero. Had a rifle that shoot well, but was always "off" at longer distances. Tried a lot of remedies, turns out my zero wasn't where I perceived it to be after my 5 shot group. More shots and help from the app, I was off by nearly an inch at 100 yards. Hits at distance now.
Well so many are not gonna take this well because this is going against a lot stuff “we” thought we knew as hand loaders. I just wish I could have saw this along time ago. So much of this makes so much sense as I look back and think about stuff I scratched my head about. Great job presenting this as I know you all put a lot of thought into how to bring this out. Thank you.All I ask is more more of this. Awesome stuff
As someone who plays with stats as part of my job, I really appreciate that you spoke without dumbing it down while at the same time making the technical aspects of error bars/regression/etc approachable to more people. I will say, though, that knowing that a shot was a "pulled shot" is a real thing, and even in High Power competition using a good old M1 Garand, I can know that I "pulled" a shot and will be able to call what zone it hit even before I get the shot called to me at 300 yards. The human component is a big factor in this equation, even when using bags, etc. Unless it is truly in a complete ransom rest where there is zero play, or a sled rifle, a human still impacts where the shot goes.
Finally! Someone puts into a video what we've been doing at the Olympic level for a long time. Minimum 30 shot groups for all lots then 60 shot groups for the top 3 lots as a final test. Thank you for this!
By the time your done testing it’s time for a new barrel lol
@@taylorbokshowan5713 not on a 22LR rimfire gun...
@@dougblessin oh ya I forgot you guys shoot 22lr ya no issues there. Test away.
That's because you're interested in the facts and not ego. Target shooters are the best people to consult, they are only interested in what gives a better score aka a smaller group. They are usually very open about what works. I always consult with other target shooters to see what's working and start from there, saves a lot of time and money.
@@redrock425 you know that confirmation bias they talked about? Your comment is a example of it.
I would love to watch a podcast with Jayden Quinlan and vortex optics, Ryan Muckenhern. Subject matter is irrelevant.
He’s been on theirs talking about terminal ballistics
@@hornady yeah, but can there be enough conversations between professionals?
,
It's been out there for some time.
@@hornady YOU CAN TAKE A 1/10TH MOA 6PPC RAIL GUN, PUT GARBAGE AMMO IN IT AND IT CAN SHOOT 2 MOA FACT!
You did a huge service to the whole community! Thank you very much for sharing this!
Our pleasure!
This has turned in to a must watch podcast.keep up the great work
Much appreciated!
An excellent podcast, thank you!
I knew 5 shot groups were insufficient but I underestimated how badly insufficient.
My take aways:
I’m not firing enough shots for a true zero.
Powder makes much more difference than I thought.
The .2 gr differences are chasing phantoms.
Excellent synopsis.
I always thought, because I heard it somewhere, that a 7 shot group was the smallest sample size to mathematically get reliable velocity data from. Now, my chrono only does 10 shot strings. What is the formula for calculating standard deviation, so I can take multiple strings, put them together and get accurate data?
@btruckno1
Shoot/Record multiple 10-shot groups, then calculate the average.?
I can appreciate the level of OCD that caused the camera to pointed in such a way to line up the edge of the laptop with the edge of the monitor. 👍
Mikes, Jayden, Seth. Great podcast. We’ve been doing similar testing as well and would like to send over data that we’ve found if interested.
Great stuff. I self discovered this visually. I shot lots and lots of .308 on various days on the same custom target. 5 and ten shot groups and kept them in a file folder. I was vexed, I had 8-9 shots under a MOA but frequently had a wider shot or two, mostly high and right. One day I am shuffling through all these targets in my stack. 40 of the same photocopied target, looking at them like one of those index card handrawn movies where you bend and release a card at a time to show motion I could see my group wasn’t small with random flyers but in fact was egg shaped with the pointy end of the egg high and right. Granted not a perfect circle like the video above but it clearly showed that my sub moa scoped M1a was really super MOA gun at 100 yards. A lot less statistics but the proof is the same visually.
200 ELD-X, 5x7 Bull Elk, 400 yards, 300 WSM, behind shoulders, no pass though, perfect mushroom! My go to forever!
FANTASTIC explanation as to why I have been chasing my zero my entire life. I will be shooting 20 shot zeros going forward. Thanks for making us better.
Great info. Your components spec out as good as any on the market. Thanks for taking the time to derive all this info, and put it in usable form. I like Jayden's comment at the end about small sample size being able to eliminate a load. I also like the comments about bullet jump.
This is the level of depth I appreciate.
Guys, AWESOME series of podcasts. I've never been a sub moa snob and you guys just confirmed that i've been on the right track. Statistics were my favorite college courses and this is why. Thanks and keep up the good work. BTW - Because of this series I'm a hornady only bullet guy for my reloads.
Why would the series lead you to purchase their bullets? In appreciation? I guess, but it's pretty well recognized that Berger and Lapua make better (both in terms of quality/consistency and performance) than Hornady . . . until you get to 750grain AMAX. Also, nothing in this video has deterred me from believing that an MOA with 30 shots is the absolute bare minimum.
@@OFFICIALUND shut up lol 😂
This is gold. And I'm thankful you've confirmed my personal methods and it mimics almost exactly my experiences and results.
I appreciate the objective information you guys always provide.
This is great getting into the nitty gritty of it and laying it out as it is! Looking forward to the app tool you mentioned to assess grouping. You guys are doing an amazing job and I truly appreciate what you are doing.
Awesome, thank you!
Fantastic podcast you guys rock. Completely ruined my load development schedule but hey ho the truth is the truth keep them coming
I appreciate your efforts to educate everyone to a sample driven approach of distribution. I am an auditor that works with statistics and samples everyday. In my mind (and in my testing which I do as a hobbyist), I try a slightly different approach. I strive to “quantify” my sample size and results by a statistical method. For example, given a population of ammunition and a given barrel, what should the sample size be to establish a quantifiable result. “Large” and “small” samples applies directly to block samples. However, if we think of a more data driven approach, what is the “appropriate” sample size to establish a 95% level of confidence, with a 5% margin of error that our results are accurate. We could further quantify the results to a group size, offering the marksman a statistical likelihood of success. To illustrate, I test “x” ammunition in “y” barrel. I can attest with a 95% chance, and within 5% margin of error that “x” ammunition will produce “y” amount of shots within a given MOA (which is to say for example…I am 95% confident with a 5% margin of error, this ammunition and in this barrel will produce x% of shots under .25 MOA, y% of shots under .5 MOA, and z% of shots within 1 MOA.) To clarify, there is a way to establish the correct sample size to provide an accurate estimate of what the barrel is able to produce with a specific ammunition.
Very interesting guys, thanks for sharing your learnings. Thoroughly enjoyed the podcast and made a few notes that'll help me work up a 300 PRC load and get it zeroed. Prioritising powder and projectile component combinations over fiddling with bullet seating depth and powder charge for a single projectile and powder combination will change how I approach finding a load that meets my expectations. Thanks again!
I watched this through another site but came here to comment. Wow, this was the most honest and accurate description/discussion of these topics I've seen, backed by data. My favorite part was "3-5 shot groups won't show you a good load, but they can show you a bad load". I'm sick of Gun Liars (writers) w 5 shot groups..."and if we throw out that flyer, it's a 0.75 MOA group!" (The "flyer" made it maybe 1 MOA...they throw out an offending data point while still showing an artificial expectation which is: it's easy to get a system which delivers 1 MOA). Liars. It's certainly doable to get 1 MOA and get it consistently, but it's not a guarantee and it's not a cake walk to get it. When we see countless articles and YT videos showing sub MOA groups as if that's the norm, even saying one can shoot 1 MOA invites ridicule because the "expectation" is that 1/2 MOA or less is what is normally shown. It's like this: "I saw a porno once and now I know that every man has a 9 inch peen!" Uh-huh...sure, bro. Anyway, I feel this information is accurate and honest and data-supported...and I've burned a lot of rounds handloading, trying to get various components to deliver consistent 1 MOA performance and I've usually gotten it (not always...I've got a 6 ARC that I'm going to have to revisit)...I've lived a lot of what was presented here. If this stuff was easy, everyone would be a 1k competitor. Thank you for the time and effort to make and post this video.
I hate, when I'm shooting a longer string though, and know I yanked one, or that a flyer IS my fault, not the gun's.
Great stuff... just wish you could have gone into slightly more detail on how you actually do load development now in light of what you learned. I saw where Miles has expanded on it a little in the comments, but a separate video or article doing a worked example or two would be awesome 👍
This is an absolutely great video. Really enjoyed it. When possible I will be trying the 30 round group and face the truth. Thank for the time and the products.
Glad it was helpful!
Guys super great job. The information you present is priceless specially for guys we don’t have the resources to spend that much on tests. Thanks a lot.
As an engineer I really appreciate this level of data analysis. It would be really interesting to show the effect the shooter has on the group size vs the locked down accuracy set up. You have a large enough data set on a known rifle system (locked down) to allow you to calculate the contribution from a specific shooter. Root Sum Squared (RSS) analysis will all you to separate the data.
My thoughts too, the most noise is generated by the shooter. Until you know how much it's a lot of guesswork.
Thank God someone has actually put in the work on this. Thank you.
Having taken undergrad statistics and research methods classes I've always had a strong intuition that basically all the testing I see people do is a lie. (Just knowing that, according to the math anyway, you need 34?, I think, data points to have a legitimate 95%CI is enough to tell you that a 3 or 5 shot group is useless. Or at least not what you think it is.) But I've never seen anyone use even pretty basic statistical analysis to show people how that works.
I'd be interested how much a 20 shot group shrinks if you throw out one. Seems like it might be a good practical measure of accuracy. Accuracy claims could start to include that qualifier of, say, 2 sigma. So instead of people claiming 1 moa (but really meaning it'll put 3-shots under 1 moa 50% of the time, but you can actually only count on 2 moa) they could say something like "this is a true X moa gun" meaning it'll put 95% of rounds inside X minutes. Period. I think it's easier for people to accept a few "fliers" in 50 rounds vs. the reality, and a cutoff has to be made somewhere.
Also, teach people what a minute of angle actually is...
I have enjoyed all this information you guys have been putting out! Thanks!!!
Thank you for this video! I've been developing an understanding of some of this stuff myself - but this video has allowed me to join the dots! Very clearly presented - awesome video!
Would love too do more shooting but we can’t get the components to do that. Really great information but not the world we live in now.
Thanks for pointing out the Law of Large Numbers in a practical setting.
Listened to the Podcast, it was that good I had to watch it. Thanks for the great content HP Team 👍
Guys … outstanding video! A must watch. You guys should write this up & publish. Loved “confirmation bias funnel”, “how many shots, how often?” “ego”, “cost of small samples”, “cheat the system”, “how bad, but not how good”. I’ve been beating my head against the wall forever, railing against the severe limitations of small group size extreme spread as a characteristic of precision. How many realize that ES is a group size dependent random variable? Not many. This video helps. But, the simplicity of ES and its visual power, in the face of costs (time, wear, components) make changing tough. The plethora of videos touting 3-shot groups to evaluate seating depth effects, ladder testing, optimal change weights, presented with such confidence by the internet gurus doesn’t help either. Keep up the good work. Five stars.
Thanks for the comment!
This podcast alone is enough to guide my reloading evolutions. Thanks much!!
Not sure why it took me so long to watch this podcast. You guys are really speaking my language. (engineering background professionally.) Makes me want to do a bunch more testing haha.
Outstanding information, gentlemen. I generally load 10 rounds somewhere near a grain or two below maximum and then fire a group over the chronograph after fouling the barrel if it was just cleaned. if the group looks decent with no signs of pressure for the application I'm done frigging around. I will generally start with the powder noted in the loading manual that generally gave the best results for a given bullet weight. But... I may start shooting 20 round groups for a solid zero after I get a good load.
Great talk!! I decided years ago with all of the rifles I have hand loaded for selecting the best bullet and powder will give you 99% of the potential of that rifle. Seating depth and small variations in powder charge may or may not yield a meaningful improvement in the remaining 1% of availible performance. As far as repeatable groups it's important to remember that the atmosphere is always changing by the minute. That's why bench rest shooters will often wait for their condition and run the gun fairly quickly. I consider a rifle sub moa if I can consistently produce 3 or 5 shot groups in isothermal air and I am fresh even if those groups move around a little between sessions. This is why you are allowed fowlers and sighters in NRA high power competition. Mirage will effect accuracy in two ways. Makes it difficult to determine aim point. And will scatter bullets especially lower bc bullet at lower velocity. Extreme spread of a small sample will invariably be closer to the SD of a large sample than the calculated SD of the small sample.
I really look forward to these podcasts.Just about to start on a 7mm-08 load with your 162 eld-m, I guess I'll give this a shot (pun intended) lol
OK, my third comment on this podcast: Of your podcasts that I have watched, this is the best one for me.
I started hunting with my dad when I was 5, which was more than 50 years ago. I never used a scope until 2008, when I bought a rifle to bring on a deer hunt, since I didn't just want to use the SMLE .303 British my dad had gifted to me (it is a capable rifle, but I wanted to put a scope on a modern rifle, so I went with a Tikka T3 lite Stainless in 270WSM, with a Leupold VX II 39-40 with a Creedmoor BDC in the reticle, pretty close to that cookie cutter setup. I had nobody to guide me, and there was no TH-cam at the time, so zero'ing involved a lot of guess work. I shot about 5 boxes of Federal Premium Vital Shok BT rounds, and became reasonably confident that I'd be able to hit a deer out to 250y, but I wasn't sure beyond that. I haven't touched my rifles in a handful of years, and I'm about to zero the rifle again, and the information you've shared here has completely changed how I'm going to do it. I had already bought a couple of boxes of ammo to get started, and I was going to do the 5 shots at 100y thing, and then test it out to 200, and then 300 (I can't test longer than that - no local range allows magnums, so I'm testing in a provincial forest - I'm in Canada) and I'll probably never shoot beyond 300 anyway - maybe out to 400, but if I can get solid data at 100, 200 and then 300, I'd be okay to try a 400y side shot on a well positioned animal (deer/elk, black bear, etc). My plan has changed from that 5 shot group and so on, to doing a 20 shot group at 100. I'm going to do five 4 shot groups on a Shoot N C target at 100y, since there are 5 red dots (one in the middle and one in each quadrant) and that will allow me to get 20 rounds into the same target, but with enough space between them to be able to clearly see most shots to be able to get accurate measurements. Once that's done, I'm going to enjoy the second box out to 200 and 300 y, and I'll confirm at 25 and 50y, just for fun. The 14 point buck I shot, after all of that practice, and all of those boxes of ammo, back in 08, was quartering towards me at about 75y, so the practice was worth it, to be accurate, since the margin for error was slim, but the point there was that it was only 75 yards, and the doe I had shot the day before was running at full speed, and was about the same distance - so DOPE matters, and that ties in with practice, assuming one wants to only hunt ethically. Love the podcasts - just getting caught up now - glad you've done so many, I have many more to listen to, many are queued up!
This reminds me of my artillery school days, when we were taking the registration class. Registration is to Field Artillery what zeroing is to riflery. We used a (IIRC) 40 round sample size for theory, and 3-6 rounds for actually registering. We talked about "assurance of validity" of the registration data being 90% that all rounds fired from data derived from a 6 round precision registration would fall within 1 Probable Error in range (indirect fire), and 99% in 2 (where 100% fall within 4 PEr). What that meant to me was my average of spottings would be on the target that often and my 155mm shells would land close enough to target for a given range that they would have effects on a thin skinned target (50m burst kill, 300 m high prob kill frag radius) if I fired the appropriate amount based on the PEr for that range and charge combination. For an abbreviated registration of 3 rounds, it's 76% and 98% respectively.
Take advantage of eager listeners to teach rudimentary statistics. So many think they are doing research when all they do is read some non-refereed print accepting only those opinions who support their beliefs.
Data, need unbiased data with adequate sample size followed by appropriate analysis for mean separation.
A great podcast!!
I have 4 medium game rifles that are 3 shot wonders, a 6.5x55, 3, 30-06's, an 03 Remington, an 03A3 Smith Carona and a Savage 110. Shooting 3 shot groups from each rifle the years they exhibit the exact behavior your data suggests.
The first 2 shots are very close and the 3rd stretch to an inch never larger, the next cold barrel 3 shot group does the same but the shots do not hit exactly as the first group. The group is very near the same size, some smaller some times but none ever bigger. However the combined groups are larger than any single group stretching to 1 3 /4". If I superimpose 10 or 15 cold bore 3 shot group I can establish a more accurate zero and improve my cold bore accuracy at all ranges.
I stopped talking to reloaders long ago because of the information they talked about today.
If you have a modicum of understanding statistics, you do this automatically.
And they are so true…
Ego plays a huge roll in people trying to accept the information given here in this podcast
Very interesting stuff and very humbling as well. I thought I was a decent precision rifle reloader. Now it's time to take my game to the next level.
This was the most enlightening podcast to date.
GREAT show! Bringing new and relevant info to the front, awesome work. I likely will slowly change my loading processes and verify with larger groups from now on.
Excellent podcast guys.
Such a breath of fresh air to hear the truth laid out for all to see & hear.
In the next podcast, I think you should drive this data home by alluding to some of the raw data you've accumulated over the years with an emphasis on the shear volume of data that you have at your disposal.
Thankyou guys & especially Jayden for telling it just as it is.
Coming from a semi conductor background, I've always known the importance of proper protocols in sample size. 3 round groups are too small, but in rifle shooting at the personal level can be costly in terms of $, time, and mental fatigue. It is understandable the natural tendency to go with less.
3 shot groups from 100-200-300-400 yards can be mentally exhausting and pricey. Just doing that you’re going through a whole box if you get a couple “fliers”. So for the regular hunter all this info could mean so little yet still great info.
I’m glad I passed my Stats final yesterday and this makes sense!
Oh man, the day after I finally got a 3-shot group a perfect triangle with holes touching at 100 yards, I stumble on this "Your groups are too small" video! Oh well, not changing my recipe. Getting older, so wider group is inevitable, and the weather will never be perfect again.
Yes, there are a multitude of factors at play here that cannot be properly controlled and are difficult to address statistically. What these chaps are saying is true for rigidly applied scientific method. And yet, I shoot 3 shot groups and identify a powder charge weight that provides groups well below 0.5 inch at 100 yards day after day and in a variety of climatic conditions. This level of accuracy extends to mid range shooting as well. There is something not quite right with this statistical analysis ( in my humble view).
I shoot once let barrel cool. More than likely I will be shooting a cold barrel at a deer or bear. Letting the barrel cool cost time. My 12.5 pinned 308 cold barrel shoots right on at 100 yards with steel cased 150 grain. Hot it's 2 inches high at 100 yards. The hotter the barrel the worse groups I see. The 125 grain varmint rounds are better than match rounds I have tried. Higher velocity. Charts also show magazine changes. If using sand bags they do flatten out after more shots. The charts can show the sand bags flattening too. I find it helpful to label mag changes. Why did that round shoot higher? I just changed magazine. The more data the better. Even ammo temp makes a difference. Keeping good notes eliminates more trial and error. I am interested in mushrooming hunting rounds. I just need to hit inner ring on paper plate. I don't usually do mag dumps. I like my barrel. 3 hours to shoot a 30rd standard mag sounds about right. Depending on outside temp. I keep my flyers. That first cold shot flyer is what I am interested in. I have to fire muzzle loader after cleaning because of clean flyer. Clean or dirty does matter. A 100 group shows?? The barrel heating up and how a hotter barrel opens up. I hunted with single shots for 20 years. A 308, 16, and 50 cal b.p. I am a budget hunter and out shoot many people. Higher cost does not equal higher quality. It shoots great with 125 grains, good with 150, and like crap with over 168 grains.
Good stuff. Im a data guy, and what you have explained on this podcast will definitely ruffle a few feathers, but in my 30+ years of handloading I have seen exactly what you guys have explained. I hate wasting components on load development, but you need a lot of data to get a good average. I hate the youtuber with 3 shot e.s. Deviation, and avg muzzle velocity. I have always based my m.v. On a 20 round string, and I am only shooting hunting rifles. People wonder how I get factory rifles shooting so good, but it involves a good break in with lots of data gathering while I am doing that. Then load development after break-in, is relatively easy. Thanks again boys!
Very informative podcasts, thank you very much. I'm glad to see that there are several Hornady bullets available in Germany. They are very reasonable priced in and the build quality is excellent. Only by visual comparison of a more expensive Geco(RWS) Star to a CX or ECX you know what I mean. The Geco bullet is an absolute joke. Hope to see more Hornady bullets available here. In .308 SST's and ELD-X bullets are not to get anywhere.
When I shoot larger groups, I stick six one inch red dots to a large target and shoot every dot 5 times with cooling in between. Then I cut the groups out, scan them and load the pictures into GRT (Gordons Reloading Tool). By giving GRT a reference distance, it automatically scales the bullet holes and the target. It's a very nice tool for reloaders and it is for free.
Great discussion. I look at cold bore kill shots at realistic distance which includes your ability, what your shooting at. Elk, Mule Deer, Antelope. I hunt with a 7MM Mag shooting 162gr ELD-X off of shooting sticks out to 700 yards. I find people get to caught up in MOA and not practical application. You can save a lot of time and money if you know your rifles capability and lots of practice. I sum up my statements like this, If your rifle shoots 2 MOA then antelope effective kill shot is around 5 hundred yards, Mule deer around 6 hundred yards and Elk would be 700 yards.
I developed a method for “scoring” five shot groups. I average the extreme spread with the best four out of five, and that gives me a score for that group. It is interesting how often that score often corresponds very closely with the mean radius for that group.
My accuracy check- first 200 rounds- put it in the deers eyeball at 100-125 yds. On full size deer target. Then experiment on out. I’m East Coast. I rarely see game past 200 yds. But I do. I think more about what that round does to the game and what’s left afterwards. Penetration and expansion. What loads get it done. Every rifle/pistol is different. Practice is what counts. But I’d have to travel for over 100 miles for a 1,000 yd. range.
Your one of the top 3 channels I watch.
It must be a pain to have to pander ţo the complainers. You could have done this episode in half the time.
You are giving great content. It has changed how I go about looking at load development.
I always though the flyer call was bunk! It is the Shooters issue.
Only way to claim that is in a great solid gun vice and then multiple shoot strings.
Then what about a slight wind, temp change, drop in barometric pressure, temp of barrel.
If you have a flyer, just own it.
@@georgejohnsmith no
58:30 You mention that the greater the distance, the larger the sample size is needed to validate the data. Do you have a suggestion/metric of required shots for say 300, 400, 500, etc. yards? If I'm understanding as a "newbie", 20 shots at 100 yards cuts your error to the ~20% range. So, 500 yards requires how many shots for 20% error to characterize your rifle?
I deal with small numbers statistics in aerospace, and confidence intervals etc. The insight here is obvious in retrospect. That’s when you know it was great work. Congratulations Hornady team.
Thank you
Good podcast. Opened my eyes to some unexpected results using 3 or 5 shot groups. Example, trying to zero rifle using 3 shot groups and having the mean POI of the next group not adjust as expected per the scope adjustment made.
Can’t tell you how many rounds I’ve fired chasing that devil…
Listened to this once upon launch. And now again. And now he's been on with Cortina's BtT podcast!.. //I'd wear a Quinlan's Corner Hornady Hat if offered. Or shirt. Just saying... 🤯🙌
Love the podcast, new to reloading.
I've watched 3 times, learn something new every time.
Keep up the great work andnthe great products
This was all great information. I’m getting ready to bring my reloading journey so I’m trying to learn all I can do I can separate fact from opinion and do the best job I can for my rifle and purposes
You really can't argue too much regarding the statistical nature of the analysis. That said, I'd love to see the same analysis (however done independently) of the quality of individual components in regard to consistency standards/measures as well as the same for off the shelf "Match" ammunition.
We all know that companies understand the shortcomings of "the egos" and market around it. Set a quality standard, put it out there and live up to it through data rather than marketing .... Not really the world we live in.
I step at 0.5 to 1.5 grains per test depending on shell volume. Looks like this is how you do it in your load manual.
This is great information. Thanks for posting these podcasts on TH-cam.
Thanks for listening
Superb presentation--many thanks. One can develop an accurate load in a quality rifle without any understanding of statistical analysis. But statistical analysis can both make the process more efficient and prevent misconceptions about the potential future accuracy of a load. With chronographs all providing a "velocity standard deviation", one does have to have some understanding of the normal distribution and statistics to appreciate WHAT that standard deviation "means" and that a sample of 20-30 tests is required for the SD to have predictive value. Handloaders really need some understanding of basic statistical analysis, as Bryan Litz has presented in his recent book. With our historical reliance on 3, 4, or 5-shot groups, handloads have long hallucinated meaningful patterns in the statistical noise of data, both with group size and "nodes" in ladder testing. Ultimately we are limited by the cost of testing and using up much of the 3000 round lifespan of many barrels. But the application of appropriate analysis can be a great aid, even without studying Box's "Statistics for Experimenters."
I've never tried it...but I'm picking up what you're putting down!!! Thanks for the video.
Hope you enjoy
I for one am experiencing a bit of cognitive dissonance after chewing on this for a few days. Your theories are making a lot of sense to me based on my own observations over the years. Perhaps I’m grateful for the explanation for the previously unexplained wild groups / “flyers” from my best rifles. I’m self aware enough to know that I’m the most likely weak link, but I also have enough experience to know when “that one wasn’t me”.
The biggest question I have is whether you guys used different brands of brass other than your own. I’m a huge huge Hornady fan and consumer but super precise brass is not what you’ve built your brand on in my opinion.
I guess I’m just fishing for something that still works to keep SD’s low and groups small despite you completely bursting my bubble. Thanks for the great info, keep ‘em coming.
I think you're referring to the Lapua elephant in the room, then of course there is Berger 😉
You've got a good point. Reality is, standard deviations are calculated for each curve. Overlaying the traditional bell curve SD over this graph is not a correct application. The regular SD curvev overlay they used wouldn't even fit over a leptoclurtic bell curve (tall & skinny) or a platicurtic (flat & wide) bell curve.
Even more to your point. I'm also a consumer of Hornady bullets & ammo... and others. I can't get smaller than 2 in groups with their whitetail ammo in my 30-06. Can't with remington core lokt either. With core lokt I get random single 2" further away (so 4" total away from center) shots about every 3-5 shots. With barnes, my biggest groups that I occasionally get so far are all consistently 2", with most of my groups being about 1", especially prior to abt 9 shots before a barrel clean and with barrel cooling after each shot.
29:35 watching this again and I missed it before. Myles is talking about if you do several separate 3, 5, or 10 shot groups and do not correlate to PoA, each group looks great but the whole overlapped pattern could be bigger. That relationship is the data being lost.
Thanks, that was very insightful and well tempered!
I’m a hobby shooter, I work a lot so when I get a free day I’m at the range with a simi stock rifle and factory ammo. I’ve determined which factory ammo my guns like the best and only have 500 yards to shoot. I shoot 308 and .223 at that distance so it’s a lot of fun. Use what you got and enjoy what you do. I enjoy listening to professionals talk about all this stuff.
Thanks for watching!
As far as practical application goes, this was in my opinion by far the most helpful episode in your ballistics series (and I've really enjoyed all of them). Here's a question that I had after listening: I've been an OCW reloader for a while and I'm curious if when testing powder charges, not for velocity but rather for POI shifts between different charge weights, does the average POI become practically equal over different charges as the sample size gets larger? I was under the impression that as the velocity changes with charge weight (now I know in a linear fashion:)) the bullet will leave the barrel at a different point in the barrel's vibration, causing changes in POI. Is this another misconception and have I been fooled by bad data and small sample sizes? And of course I'm looking for practical application, without getting caught up in the minutia of theory. Thanks and keep up the good work!
The hardest thing for me to find is what a bullet will do on game at different velocities. that aren't just one bullet test. This would make some great promotional material and a podcast. Great info as always.
well im going to shoot a 50shot group... in a 100yd tube...
My takeaway from this as a hunter is, that if I shoot a 5 shot group, I can take that average group size and increase it by 50% and that should be very close to what I would get in a large 50 shot group or (x .75 for a 100 shot group)
Thus a with a 1”, 5 shots avg then I could with some confidence say I have a 1.5 inch gun with that combination over 50 shots …
… and if I used your 100 shot graph then I could confidently but sheepishly admit that I have a 1.75 inch or better "all day long, any day" gun & load combo
I will endeavor to shoot 5 shot groups and accept that my REAL accuracy with that 5 shot group avg with that rifle load combination over time is actually 50% larger than my average 5 shot group.
If I want to hunt a 4 MOA target sized animal then I need a 5 shot group that shoot better than 2MOA 5 shot groups at whatever my limiting distance is for that animal.
I failed statistics but think I can understand and use someone else's explained statistical data.
If you read these and this is wrong please correct me.
What an amazing video. Cant argue the data. Gonna make me look at load development differently for sure!
This was a stellar podcast. Worth watching and referring to multiple times. You really went to the essence of what we do as reloaders and shooters interested in precision. Thank you so much.
2 questions on what you did:
1) What specific method did you use to measure the distance to the lands?
2) When reloading, did you use base to ogive distance or did you use COAL?
This is why people who really know rifles and precision shooting say that any rifle that’s truly sub minute is a real gem.
The one that always makes me shake my head are the claims of new extreme range records. The one where they shoot 100s of shots and eventually one lands on a house sized target 4 miles away. My response is always "Nice. Do it again with the next shot, then I'll be impressed".
I definitely think those are cool but they are absolutely novelty records
Yes, a lot of shooters can volley enough rounds down range to hit close to an intended target given enough time and rounds, but it's never repeatable
This was funny because anyone who follows BR already understood that most people don't have a .25' gun. If they did, they could win the BR competitions.
My favorite group pattern on a public range is a smiley face. If you want to hear people cuss, this is the most fun. Shooting 1/8" groups gets boring, so golf balls are fun targets. Usually large amounts of data is easier to analyze with RMS as the environmental changes need to be factored out to find a good load. If you could shoot underground and draw a vacuum from the muzzle to the target, you could find the perfect load, then add a little to handle a real world environment. I found to shoot while it is raining to get the best results as the humidity, sunlight, temp, and wind remains the same.
Great information. The chasing my zero part hit solid for me. I’ve done it more than once. Never did figure out the issue until just now
Glad it helped!
I have thoroughly enjoyed this series. Guys, thank you!
Glad you enjoy it!
You guys have done such a good job explaining the concepts, that I spent last night changing my next planned load test paperwork from 10 incremental 3-shot groups to 3 drastic 10-shot groups. Now I'm going to hold off on all tests until I catch up on the series.
Glad it was helpful!
Amazing stuff guys!! Definitely going to watch this one again. Just like I do with all the others. Its the only way my brain can absorb the information.
Awesome! Thank you!
First, a lot of valid points are made in this video.
I'm also battling myself in making this further comment. It's coming from my experience in undergrad and grad level statistics courses (no desire for self praise or bragging or proporting academics) and my caution because I don't want to come across as a know-it-all by making the comment. I do suspect some might find it helpful.
The thing that I'm seeing though is your generalizing statistical analysis and not accounting for some statistical factors that exist in your graph, which I can see will skew the interpretation and alter perceived facts/outcomes.
There's too much to get I to here, but here's a few that affect the outcome. First, there are 3 types of averages - mean, median and mode. Mean is kind of the one most people are used to with bell curves. What you used appears to possibly be the mean, but the median appears close enough to be hard to tell here.
A big thing I'm seeing here is there are so many modes (clusters of the same number that occur) that those will skew/shift the results and accurate interpretation.
On the right side the modes skew it higher and the left lower, so typical bell curve averages are reduced in how normal average interpretations can apply.
The typical standard deviation scale can virtually be placed over any curve and get similar equivalents, but doing so on skewed curves can lead to incorrect interpretation of the results.
When people are testing loads, they're looking to see if there's any advantages of combinations of powder loads, primers, bullets, etc. Which is basically trying to see if there's any significance of one combination vs another. In such a case, more likely the statistical measures to be sought is high modes vs low modes. Higher modes are going to be a sign of a higher performing loads (more desirable in shooting). The left side of your curve having more lower modes, look like they are lower performers vs the right. This appears to be consistent with your histogram showing an upward trend as powder loads increased, which is also to be expected.
In true statistical analysis you'd actually want to first run standard deviation math to identify what the actual standard deviation numbers would apply to this curve. Then run the appropriate statistical math tests (i.e., chi square, anova, canova, etc depending on some factors) comparing the modes to see if any mode is found to be statistically significantly better. If not, statistically significant, the improvements would be considered to occur randomly and not reliable enough to depend on. On your graph, it looks to me like it's right skewed, which which will also influence your mathematically correct standard deviations. The regular standard deviations you're using apply to regular bell curves, which you don't really have.
To me, looking at face value of the modes, I think there's a chance that some of those on the right might actually be statistically better performers. If so, you would want to load those more, which will all be more likely better performers (have better averages than your original bell curve).
Further. Shooting a barrel out will give you a census (not sample) on that barrels performance. Which is not very useful because the barrels not usable anymore, unless you're comparing manufacturer gun performances.
The purpose of "samples" would be to get an idea of how a barrel will likely perform on generally. How to you get a credible sample size per barrel, you'd need to know the barrel life capabilities and use a % of that which would be considered reliable. For example. For a total population of 1000, there's a standard % that usually, after that #, the results start to repeat themselves. So the sample size is considered reliable, though some outliers can still exist but likely wouldn't impact the results.
I say all this in effort to help people who really want to find better averages in their loads. There's just some stuff missing and misunderstood here that a more accurate understanding of might be of interest to some.
Hope it helps.
Happy shooting!
I hope they look into all this. Send them an email, I'm curious what would be learned from the kind of analysis you're talking about
Great info. Out of curiosity with your testing, did you use factory ammo or were you reloading these rounds. The reason I ask is because after the first episode you did regarding sample sizes I did try the 20 shot group method. Wasn't surprised to see the group measured 3/8 inch as this is a known load. If I disassemble the rounds I loaded, the powder charges are accurate to within .02 grains measured on an FX-120i scale.
I had two boxes of leftover factory ammo for a 6.5 CM that I decided to disassemble and see what the charge weights were and found charge weights were not at all consistent. Berger factory ammo had a spread from 40.1 to 41.9 grains and Hornady ELD match went from 39.4 to 42.3 grains.
So my question is, if factory match ammo was used in these tests, how accurate are the results if the variation of said ammo can be around 2 grains of powder difference?
Overall awesome podcast guys. Please keep them coming. I just had to ask as I have seen questionable results with factory ammo.
You kinda can't deny the bench rest guys do get the variables skimmed down when they do ladder test and seating depth test, they really do get the guns dialed in, but then they also use tuners which is harmonic witch craft.
Wow!!! How much time have I been wasting over the years!
Question - if you are just plumping for a bullet and powder, how do you choose the powder weight? Just go to max minus 5%?
Thanks for a great podcast which will undoubtedly generate a huge amount of discussion in the forums :)
Pretty much. I typically drop 0.5-1.5gr from book max (depends on the cartridge, 1.5gr in a 5.56 is quite a bigger change than 1.5gr in a .300 PRC). Just looking for something that will be powerful enough to cycle the firearm (if that's a concern), to get enough velocity to achieve my range/terminal performance criteria, and that's going to be safe.
From what they said in this podcast it’s not relevant, faster shoots flatter and kicks harder so pick accordingly.
Eye fatigue, shooter fatigue, fine motor control, hot barrel, etc...
Miles, somewhere in one of the podcasts you were on you mentioned your PRS ARC loads. Wondered if you could provide them again here. I’m looking for a point of reference other than the book. I have been shooting 108 ELDM with 28.6 of LVR. Also ran a few loads of Varget 27 grains with the same bullet, to much pressure.
Running this in a Blueprinted Rem 700 action with a 24” barrel, diamondback special trigger.
Because there are so many variables in ballistics it does leave a lot of room for myth, rumour and superstition. Bullet, case, primer, powder. Then you have charge weight and seating depth, distance to lands. Add to that the rifle, the shooter, the scope, atmospheric conditions. Almost infinite combinations.
My take experience is there are lots of good combinations and some terrible ones. As with shooting to improve your score, eliminate the worst shots and eventually you'll be left with them all in the X. Start with a quality rifle and scope all set up correctly. Choose good known components to start with. Load to the velocity you require for your needs. I'd start with 10% below max on a fullbore rifle at SAAMI specs and then go from there making one change at a time.
This is amazing info, thanks for doing this! I stumbled across this podcast by accident and I'm glad I did. I would love to hear more on the topic you did about why rounds don't all go through the same hole, i.e. what causes dispersion. Especially more examples of what matters and what doesn't etc.
Episode 57 is all about dispersion!
Question. Will your tool you hinted at for measuring group size be able to aggregate multiple groups into one aggregate group? If not, everything you've said in this video has missed the ears of your software engineers
I was out chasing my tail last weekend. This was very helpful. Thank you.
Glad it helped!
This is pure gold! Keep the content coming! 🔥🔥🔥
the info from you guys is gold.
now how much care was taken in the reloading process doing these tests? Were bullets sorted for length? Was seating depth perfect on all cartridges? Did all of the cases have the same headspace? Or was it all factory ammo?